opensource/ai-edu
3
0
Fork 0
mirror of https://github.com/microsoft/ai-edu.git synced 2026-09-21 04:26:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

Zixuan/dict (#507)

Browse Source 
* fix path

* fix path
This commit is contained in:
Zixuan Chen
2020-05-27 10:38:43 +08:00
committed by GitHub
parent e2f5a76c67
commit 5361c16bc8
1 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
B-实践案例/B13-AI对联生成案例/docs/fairseq.md
+6 -4
View File
@@ -199,14 +199,16 @@ fairseq-interactive ${PREPROCESSED_DATA_DIR} --path ${MODEL_SAVE_DIR}/checkpoint
```
2. 读入模型文件
第一个参数为checkpoints所在目录，`checkpoint_file`为需要读入的checkpoint的文件名，`data_name_or_path`为字典文件所在的目录。
我们将模型生成的checkpoint文件拷贝到`./checkpoints`目录下，将预处理阶段生成的两份字典文件（`dict.up.txt`和`dict.down.txt`）拷贝到`./checkpoints/dict`目录下。按如下代码即可读入文件：
```
model = LSTMModel.from_pretrained('./checkpoints',\
checkpoint_file='checkpoint_best.pt',\
data_name_or_path="DICT_PATH")
data_name_or_path="./dict")
```
第一个参数为checkpoint文件所在目录，`checkpoint_file`为需要读入的checkpoint的文件名，`data_name_or_path`为字典文件所在的目录。需要注意的是，字典文件目录将以第一个参数为根目录，若路径有误会出现报错信息：`AttributeError: 'NoneType' object has no attribute 'split'`。
3. 推理
此处需要注意输入的文字之间需用空格隔开。
@@ -233,7 +235,7 @@ fairseq-interactive ${PREPROCESSED_DATA_DIR} --path ${MODEL_SAVE_DIR}/checkpoint
model = LSTMModel.from_pretrained('./checkpoints',\
checkpoint_file='checkpoint_best.pt',\
data_name_or_path="DICT_PATH") # 读入模型
data_name_or_path="./dict") # 读入模型
app = Flask(__name__)